CN117853407A - Highway disease prediction and maintenance method based on attention mechanism DCGAN time sequence model - Google Patents

Highway disease prediction and maintenance method based on attention mechanism DCGAN time sequence model Download PDF

Info

Publication number
CN117853407A
CN117853407A CN202311619328.7A CN202311619328A CN117853407A CN 117853407 A CN117853407 A CN 117853407A CN 202311619328 A CN202311619328 A CN 202311619328A CN 117853407 A CN117853407 A CN 117853407A
Authority
CN
China
Prior art keywords
dcgan
attention
time
model
attention mechanism
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311619328.7A
Other languages
Chinese (zh)
Inventor
李明
王琳
葛恒志
吕方惠
夏敏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Clp Hongxin Information Technology Co ltd
Original Assignee
Clp Hongxin Information Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Clp Hongxin Information Technology Co ltd filed Critical Clp Hongxin Information Technology Co ltd
Priority to CN202311619328.7A priority Critical patent/CN117853407A/en
Publication of CN117853407A publication Critical patent/CN117853407A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0475—Generative networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/094—Adversarial learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a highway disease prediction and maintenance method based on a attention mechanism DCGAN time sequence model, which comprises the following steps: constructing an attention mechanism algorithm model, defining an input sequence in the attention mechanism algorithm model by utilizing collected historical highway video, image or text characteristic data, and outputting an attention weighting matrix through the attention mechanism algorithm model; calculating by using the attention weighting matrix, and outputting a time input sequence; constructing a DCGAN time sequence model, and performing countermeasure training according to the time input sequence through the DCGAN time sequence model; and obtaining a predicted evaluation result according to the time input sequence corresponding to the road section to be tested by using the optimized DCGAN time sequence model, and making a maintenance plan according to the evaluation result. The invention makes up the defects of the traditional highway disease prediction method, can identify important time nodes in the highway disease image sequence, capture important correlations between images at different time points and predict future disease development conditions.

Description

Highway disease prediction and maintenance method based on attention mechanism DCGAN time sequence model
Technical Field
The invention relates to a highway disease prediction and maintenance method, in particular to a highway disease prediction and maintenance method based on a attention mechanism DCGAN time sequence model.
Background
Along with the gradual perfection of a highway network and the increase of traffic, the problems of highway diseases such as cracks, pits, collapse and the like are gradually revealed, so that traffic accidents can be caused, the running safety of vehicles is threatened greatly, and the traffic management department consumes a great deal of manpower, material resources and financial resources to detect, repair and maintain the highway diseases. In this context, roads are also gradually shifted from construction to operation and maintenance management. And the road disease prediction is carried out by integrating factors such as climate, traffic volume, road structure, road current situation and the like, so that potential disease risks can be found in time, a maintenance plan is formulated in advance, resources are reasonably arranged, the maintenance cost is reduced, the maintenance efficiency is improved, the occurrence of traffic accidents is further reduced, and the service life of the road is prolonged. The accurate prediction of highway diseases is beneficial to the traffic management department to better understand the distribution and development trend planning of highway diseases and further optimize the traffic transportation network.
The existing patent methods for predicting the road surface diseases are developed from two aspects, namely, on the equipment acquisition method, road panoramic images, position information and the like are obtained through panoramic image acquisition equipment, and then the evolution trend of the road surface diseases is analyzed; secondly, on the deep learning method, a disease data training set is constructed through a deep convolutional neural network to classify and evaluate the damage condition of multiple types of road surface diseases. However, due to different materials, structures, environments and traffic flows of different roads, the result shown by disease prediction by simply analyzing the road surface images is not accurate, and the establishment of a maintenance plan can be further affected. And the prediction precision can be improved by considering both time sequences and multiple influencing factors, so that a better support is provided for developing a maintenance plan.
However, the road disease prediction needs to comprehensively consider the multi-factor characteristics and the time sequence, so that a road disease prediction and maintenance method based on a attention mechanism DCGAN (deep convolution generation countermeasure network) time sequence model is needed to solve the problem that the result displayed by simply analyzing the road surface image to predict the disease is inaccurate due to different materials, structures, environments and traffic flows of different road surfaces.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a highway disease prediction and maintenance method based on a attention mechanism DCGAN time sequence model, so as to solve the problem that the result displayed by simply analyzing road surface images to predict the disease is inaccurate due to different materials, structures, environments and traffic flows of different road surfaces.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
a highway disease prediction and maintenance method based on a attention mechanism DCGAN time sequence model is characterized by comprising the following steps:
constructing an attention mechanism algorithm model, defining an input sequence in the attention mechanism algorithm model by utilizing collected historical highway video, image or text feature data, calculating according to the defined input sequence by the attention mechanism algorithm model, and outputting an attention weighting matrix;
calculating the collected historical highway video, image or text feature data and an attention weighting matrix, and outputting a time input sequence with attention weights; constructing a DCGAN time sequence model, performing countermeasure training according to a time input sequence through the DCGAN time sequence model, and obtaining an optimized DCGAN time sequence model;
defining an input sequence of an attention mechanism algorithm model by utilizing the collected video, image or text characteristic data of the road section to be tested, and outputting a corresponding attention weighting matrix through the attention mechanism algorithm model; and calculating the video, image or text characteristic data of the road section to be tested and the corresponding attention weighting matrix, outputting a corresponding time input sequence, obtaining a predicted evaluation result by using the optimized DCGAN time sequence model according to the corresponding time input sequence, and making a maintenance plan according to the evaluation result.
In order to optimize the technical scheme, the specific measures adopted further comprise:
further, the video, image or text characteristic data acquire the characteristic data of the road condition through the vehicle-mounted inspection equipment.
Further, the characteristic data comprise road surface damage degree, road materials, road structures, temperature and humidity changes and traffic flow.
Further, the collected real-time video, image or text characteristic data of the road history or the road section to be detected are segmented according to time steps, and the images are automatically marked for preprocessing.
Further, the algorithm model of passing through the attention mechanism calculates according to the defined input sequence, and outputs an attention weighting matrix, specifically: defining matrix parameters of an attention mechanism algorithm model, calculating an attention score by using an input sequence and the matrix parameters, dividing the attention score by a scaling factor, and obtaining a scaled attention score; and applying the scaled attention score to a softmax function to obtain normalized attention weight distribution, normalizing the attention weight distribution to ensure that the sum of all weights is equal to 1, and carrying out weighted summation operation on the normalized attention weight distribution and an input sequence to obtain a final attention weighting matrix.
Further, the matrix parameters include a weight matrix, a query matrix, a key matrix, and a value matrix.
Further, after the attention mechanism algorithm model is calculated according to the defined input sequence and the attention weighting matrix is output, historical road video, image or text feature data is used as a data source and is divided into a training set and a testing set according to the proportion of 8:2, and the constructed attention mechanism algorithm model is trained and checked.
Further, the constructing a DCGAN time sequence model, performing countermeasure training according to a time input sequence through the DCGAN time sequence model, and obtaining an optimized DCGAN time sequence model specifically includes: constructing a DCGAN network structure, defining a generator and a discriminator structure, wherein the generator adopts a convolution layer and an attention mechanism, the discriminator adopts the convolution layer, a specified loss function is Wasserstein distance, and the optimizer is RMSprop; the cyclic training generator generates more real samples, and the training discriminator distinguishes the real samples from the false samples; and comparing the visualized generated sample with the real sample, adjusting an attention mechanism or a network structure according to the generated effect, and continuing the countermeasure training until convergence so as to obtain an optimized DCGAN time sequence model.
Further, the obtaining a predicted evaluation result by using the optimized DCGAN time sequence model according to the corresponding time input sequence specifically includes: the method comprises the steps of inputting a time input sequence into an optimized DCGAN time sequence model, extracting spatial relations among features by a generator of the DCGAN time sequence model, strengthening learning ability of the model to important features by using a attention mechanism, performing deep learning on input data by using the generator of the DCGAN time sequence model, and predicting future conditions of a road by using the ability of generating countermeasure training.
Further, the making of the maintenance plan according to the evaluation result specifically includes: and identifying which road sections are damaged to exceed the standard within a certain time preset in the future, and evaluating the maintenance priority of different road sections.
The beneficial effects of the invention are as follows:
the invention can make up the defect that the traditional highway disease prediction method has single input data and does not consider the time sequence of the image, and can identify important time nodes in the highway disease image sequence; important association between images at different time points can be captured, so that future disease development conditions can be predicted; the method can generate high-quality highway disease images, is beneficial to supplementing actual monitoring data and expanding training sample size; the method can combine the generation task and the prediction task, and the two tasks can mutually promote, for example, the generated image is utilized to increase the data volume of the prediction task, and meanwhile, the attention mechanism and the time sequence information are adopted, so that the change rule of the highway diseases along with time can be better analyzed; the DCGAN time sequence model is combined with the convolutional neural network, has strong processing capacity on image data, can predict road disease conditions at a plurality of time points in the future, and provides decision support for road maintenance. The prediction result is more accurate and reliable, which is favorable for taking prevention and control measures in advance and reducing the economic loss caused by highway diseases.
Drawings
Fig. 1 is a flowchart of a highway disease prediction and maintenance method based on a DCGAN time series model of an attention mechanism.
FIG. 2 is a flow chart of an attention mechanism algorithm model stage of a highway disease prediction and maintenance method based on an attention mechanism DCGAN time sequence model;
fig. 3 is a flowchart of a DCGAN time series model phase of a highway disease prediction and maintenance method based on a attention mechanism DCGAN time series model according to the present invention.
Description of the embodiments
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings of the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
As shown in fig. 1, the highway disease prediction and maintenance method based on a attention mechanism DCGAN (deep convolution generating countermeasure network) time sequence model according to the embodiment of the invention includes the following steps:
constructing an attention mechanism algorithm model, preprocessing, such as standardization, denoising and the like, by utilizing collected historical feature data of road video, images or texts, and then defining an input sequence in the attention mechanism algorithm model, calculating according to the defined input sequence through the attention mechanism algorithm model, and outputting an attention weighting matrix;
preprocessing collected historical highway video, image or text characteristic data, calculating with an attention weighting matrix, and outputting a time input sequence with attention weight; the important characteristic ordering of the influence factors is given by using an attention mechanism algorithm model, and the characteristic data segmented in time sequence are combined to be used as a data input source of a DCGAN time sequence model; constructing a DCGAN time sequence model, performing countermeasure training according to a time input sequence through the DCGAN time sequence model, and obtaining an optimized DCGAN time sequence model;
defining an input sequence of an attention mechanism algorithm model by utilizing the collected video, image or text characteristic data of the preprocessed road section to be detected, and outputting a corresponding attention weighting matrix through the attention mechanism algorithm model; and calculating the video, image or text characteristic data of the road section to be tested and the corresponding attention weighting matrix, outputting a corresponding time input sequence, obtaining a predicted evaluation result by using the optimized DCGAN time sequence model according to the corresponding time input sequence, and making a maintenance plan according to the evaluation result.
The method comprises a attention mechanism algorithm feature ordering stage, a DCGAN time sequence model training stage and a highway disease prediction stage, wherein the attention mechanism algorithm is firstly utilized to obtain the weight of a multi-factor feature input position through the steps of defining an input sequence, matrix parameters, calculating a scaled attention score, performing softmax function operation and the like. And secondly, combining the preprocessing data with an important feature sequence to serve as the data input dimension of the DCGAN time sequence model training, and finally realizing the prediction of the evolution of the highway diseases through model training and testing.
The video, image or text feature data in the above embodiment may be used to collect feature data related to road conditions in real time through a vehicle-mounted inspection device (such as a camera, a sensor, etc.).
The characteristic data comprise road surface damage degree, road materials, road structures, temperature and humidity changes and vehicle flow so as to define dimensions, length, data structures and the like of an input sequence.
In the above embodiment, the collected feature data such as real-time video, image or text of the road history or the road section to be tested is segmented according to time steps, and is automatically marked by the image, that is, the image data is automatically marked by a LabelImg tool and the like, and a data set to be input is generated, so as to perform preprocessing, and the data set is used as a data source of the attention mechanism algorithm.
In the above embodiment, as shown in fig. 2, the attention mechanism algorithm model calculates according to the defined input sequence, and outputs an attention weighting matrix, specifically: defining matrix parameters of an attention mechanism algorithm model, calculating an attention score by using an input sequence and the matrix parameters, dividing the attention score by a scaling factor, and obtaining a scaled attention score; and applying the scaled attention score to a softmax function to obtain normalized attention weight distribution, normalizing the attention weight distribution to ensure that the sum of all weights is equal to 1, and carrying out weighted summation operation on the normalized attention weight distribution and an input sequence to obtain a final attention weighting matrix.
For example, assuming an input sequence of X, comprising N elements, each element consisting of a feature vector or feature matrix; firstly, multiplying an input sequence X by a matrix parameter W to obtain an attention score A, and dividing A by a scaling factor, such as the dimension of the attention score A, wherein the scaling factor can be the dimension of the attention score or other constants; the scaled attention score a was applied to the softmax function resulting in a normalized attention weight distribution.
The Matrix parameters include a Weight Matrix (Weight Matrix), a Query Matrix (Query Matrix), a Key Matrix (Key Matrix), and a Value Matrix (Value Matrix), among others, to calculate the attention score by using the Matrix parameters.
In the above embodiment, after the attention mechanism algorithm model is calculated according to the defined input sequence and the attention weighting matrix is output, the historical feature data such as highway video, image or text is used as the data source, and is divided into the training set and the test set according to the ratio of 8:2, and the constructed attention mechanism algorithm model is trained and checked.
As shown in fig. 3, in the above embodiment, a DCGAN time series model is constructed, and countermeasure training is performed according to a time input sequence by using the DCGAN time series model, and an optimized DCGAN time series model is obtained, which specifically includes: constructing a DCGAN network structure, defining a generator and a discriminator structure, wherein the generator adopts a convolution layer and an attention mechanism, the discriminator adopts the convolution layer, a compiling model is used for model compiling, a loss function is designated as a Wasserstein distance, and an optimizer is RMSprop; the cyclic training generator generates more real samples, and the training discriminator distinguishes the real samples from the false samples; and comparing the visualized generated sample with the real sample, adjusting an attention mechanism or a network structure according to the generated effect, and continuing the countermeasure training until convergence so as to obtain an optimized DCGAN time sequence model.
In this embodiment, in terms of model compiling and training, a Mean Square Error (MSE) is used as a loss function, an average absolute error (MAE) is used as an evaluation index, an Adam optimizer is used to update parameters in model compiling and pit training, a loss value and an evaluation index of a model on a test set are calculated through an MSE method, and methods such as adjusting a value of a super-parameter, increasing the layer number of the model and the like can be tried to repeatedly adjust and train the model.
The method comprises the steps of obtaining a predicted evaluation result according to a corresponding time input sequence by using an optimized DCGAN time sequence model, and specifically comprises the following steps: the time input sequence is input into an optimized DCGAN time sequence model, a generator of the DCGAN time sequence model adopts a convolutional neural network to extract the spatial relationship among the features, the learning capacity of the model to important features is enhanced by using an attention mechanism, the generator of the DCGAN time sequence model carries out deep learning on newly input data, and the future condition of a road is predicted by using the capacity of generating countermeasure training.
The maintenance plan is formulated according to the evaluation result, and specifically comprises the following steps: and identifying which road sections are out of standard in a certain time preset in the future, and evaluating the maintenance priority of different road sections, wherein the maintenance is required in time.
The invention can make up the defects that the traditional highway disease prediction method has single input data and does not consider the time sequence of the images, and can identify important time nodes such as key frames in the highway disease image sequence, which is important for predicting the development trend of the disease. The model can capture important relations between images at different time points, such as relation of occurrence and expansion of diseases, which is beneficial to predicting future disease development. High-quality highway disease images can be generated, actual monitoring data can be supplemented, and training sample size can be expanded. The model combines the generation task and the prediction task, and the two tasks can mutually promote, for example, the generated image is utilized to increase the data volume of the prediction task, and meanwhile, the attention mechanism and the time sequence information are adopted, so that the change rule of the highway diseases along with time can be better analyzed. The DCGAN model is combined with the convolutional neural network, has strong processing capacity on image data, can predict road disease conditions at a plurality of time points in the future, and provides decision support for road maintenance. The prediction result is more accurate and reliable, which is favorable for taking prevention and control measures in advance and reducing the economic loss caused by highway diseases.
The scheme of the invention fully considers the influence of factors such as road materials, climate conditions, road structures and the like on road diseases on the basis of time sequence image analysis and detection, and greatly improves the precision of disease prediction of different road segments, thereby providing guidance for the establishment of maintenance plans. The traditional disease prediction method has the advantages that the data input is single, only the collected images are marked, identified and predicted, the importance ranking is carried out on the factors influencing the road current situation, the climate condition, the road service time and the like of the occurrence of the highway diseases through the attention mechanism algorithm, the importance characteristics are selected as input data sources, and the prediction precision of the highway diseases is greatly improved. The deep convolution generation countermeasure network (DCGAN) time sequence model utilized by the method utilizes the deep convolution network to replace a fully connected network, can segment characteristic data according to time step length, and carries out convolution operation on input data, so that the highway disease prediction precision is further improved.
In another embodiment, the invention provides a computer readable storage medium storing a computer program for causing a computer to execute a highway disease prediction and maintenance method based on the attention mechanism DCGAN time series model.
In another embodiment, the present invention provides an electronic device, including: the system comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the highway disease prediction and maintenance method based on the attention mechanism DCGAN time sequence model when executing the computer program.
In the embodiments disclosed herein, a computer storage medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The computer storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a computer storage medium would include one or more wire-based electrical connections, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above examples, and all technical solutions belonging to the concept of the present invention belong to the protection scope of the present invention. It should be noted that modifications and adaptations to the invention without departing from the principles thereof are intended to be within the scope of the invention as set forth in the following claims.

Claims (10)

1. A highway disease prediction and maintenance method based on a attention mechanism DCGAN time sequence model is characterized by comprising the following steps:
constructing an attention mechanism algorithm model, defining an input sequence in the attention mechanism algorithm model by utilizing collected historical highway video, image or text feature data, calculating according to the defined input sequence by the attention mechanism algorithm model, and outputting an attention weighting matrix;
calculating the collected historical highway video, image or text feature data and an attention weighting matrix, and outputting a time input sequence with attention weights; constructing a DCGAN time sequence model, performing countermeasure training according to a time input sequence through the DCGAN time sequence model, and obtaining an optimized DCGAN time sequence model;
defining an input sequence of an attention mechanism algorithm model by utilizing the collected video, image or text characteristic data of the road section to be tested, and outputting a corresponding attention weighting matrix through the attention mechanism algorithm model; and calculating the video, image or text characteristic data of the road section to be tested and the corresponding attention weighting matrix, outputting a corresponding time input sequence, obtaining a predicted evaluation result by using the optimized DCGAN time sequence model according to the corresponding time input sequence, and making a maintenance plan according to the evaluation result.
2. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time sequence model according to claim 1, wherein the method is characterized in that: and the video, image or text characteristic data acquire the characteristic data of the road condition through the vehicle-mounted inspection equipment.
3. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time series model according to claim 2, wherein the method is characterized in that: the characteristic data comprise road surface damage degree, road materials, road structures, temperature and humidity changes and traffic flow.
4. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time sequence model according to claim 1, wherein the method is characterized in that: and cutting the collected real-time video, image or text characteristic data of the road history or the road section to be tested according to time steps, and automatically marking through the image so as to perform preprocessing.
5. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time series model according to claim 1, wherein the attention mechanism algorithm model calculates according to the defined input sequence and outputs an attention weighting matrix, specifically: defining matrix parameters of an attention mechanism algorithm model, calculating an attention score by using an input sequence and the matrix parameters, dividing the attention score by a scaling factor, and obtaining a scaled attention score; and applying the scaled attention score to a softmax function to obtain normalized attention weight distribution, normalizing the attention weight distribution to ensure that the sum of all weights is equal to 1, and carrying out weighted summation operation on the normalized attention weight distribution and an input sequence to obtain a final attention weighting matrix.
6. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time series model according to claim 5, wherein the method is characterized by comprising the following steps of: the matrix parameters include a weight matrix, a query matrix, a key matrix, and a value matrix.
7. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time sequence model according to claim 1, wherein after calculation is performed according to a defined input sequence through an attention mechanism algorithm model and an attention weighting matrix is output, historical highway video, image or text characteristic data is used as a data source and is divided into a training set and a testing set according to the proportion of 8:2, and the constructed attention mechanism algorithm model is trained and checked.
8. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time series model according to claim 1, wherein the constructing the DCGAN time series model, performing countermeasure training according to a time input sequence through the DCGAN time series model, and obtaining an optimized DCGAN time series model specifically comprises: constructing a DCGAN network structure, defining a generator and a discriminator structure, wherein the generator adopts a convolution layer and an attention mechanism, the discriminator adopts the convolution layer, a specified loss function is Wasserstein distance, and the optimizer is RMSprop; the cyclic training generator generates more real samples, and the training discriminator distinguishes the real samples from the false samples; and comparing the visualized generated sample with the real sample, adjusting an attention mechanism or a network structure according to the generated effect, and continuing the countermeasure training until convergence so as to obtain an optimized DCGAN time sequence model.
9. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time series model according to claim 8, wherein the predicting evaluation result is obtained by using the optimized DCGAN time series model according to the corresponding time input sequence, specifically comprising: the method comprises the steps of inputting a time input sequence into an optimized DCGAN time sequence model, extracting spatial relations among features by a generator of the DCGAN time sequence model, strengthening learning ability of the model to important features by using a attention mechanism, performing deep learning on input data by using the generator of the DCGAN time sequence model, and predicting future conditions of a road by using the ability of generating countermeasure training.
10. The highway disease prediction and maintenance method based on the attention mechanism DCGAN time series model according to claim 8, wherein the maintenance plan is formulated according to the evaluation result, specifically comprising: and identifying which road sections are damaged to exceed the standard within a certain time preset in the future, and evaluating the maintenance priority of different road sections.
CN202311619328.7A 2023-11-30 2023-11-30 Highway disease prediction and maintenance method based on attention mechanism DCGAN time sequence model Pending CN117853407A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311619328.7A CN117853407A (en) 2023-11-30 2023-11-30 Highway disease prediction and maintenance method based on attention mechanism DCGAN time sequence model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311619328.7A CN117853407A (en) 2023-11-30 2023-11-30 Highway disease prediction and maintenance method based on attention mechanism DCGAN time sequence model

Publications (1)

Publication Number Publication Date
CN117853407A true CN117853407A (en) 2024-04-09

Family

ID=90542431

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311619328.7A Pending CN117853407A (en) 2023-11-30 2023-11-30 Highway disease prediction and maintenance method based on attention mechanism DCGAN time sequence model

Country Status (1)

Country Link
CN (1) CN117853407A (en)

Similar Documents

Publication Publication Date Title
CN110647539B (en) Prediction method and system for vehicle faults
CN117495210B (en) Highway concrete construction quality management system
EP3671201B1 (en) An improved method for evaluating pipe condition
Nik et al. Hybrid PSO and GA approach for optimizing surveyed asphalt pavement inspection units in massive network
CN111143932A (en) Bridge health state assessment method, device, system and equipment
CN115471625A (en) Cloud robot platform big data intelligent decision method and system
CN112183906B (en) Machine room environment prediction method and system based on multi-model combined model
CN117191147A (en) Flood discharge dam water level monitoring and early warning method and system
CN114611372A (en) Industrial equipment health prediction method based on Internet of things edge calculation
CN117037421A (en) Rain-falling landslide hidden danger group meteorological risk early warning method, equipment and storage medium
CN117291781A (en) Sudden water pollution tracing method, equipment and medium
CN112115817B (en) Remote sensing image road track correctness checking method and device based on deep learning
Wu et al. Optimization of unmanned aerial vehicle inspection strategy for infrastructure based on model-enabled diagnostics and prognostics
Heydarian et al. Automated benchmarking and monitoring of an earthmoving operation's carbon footprint using video cameras and a greenhouse gas estimation model
CN115063337A (en) Intelligent maintenance decision-making method and device for buried pipeline
CN117422990A (en) Bridge structure classification and evaluation method based on machine learning
CN117251722A (en) Intelligent traffic management system based on big data
CN117853407A (en) Highway disease prediction and maintenance method based on attention mechanism DCGAN time sequence model
CN113516179B (en) Method and system for identifying water leakage performance of underground infrastructure
Fuchsberger et al. A correlation network model for structural health monitoring and analyzing safety issues in civil infrastructures
Kim et al. Seismic Performance Management of Aging Road Facilities in Korea: Part 2− Decision-making Support Technology and Its Application
CN118015839B (en) Expressway road domain risk prediction method and device
Zhou et al. A Modelling Tool for Rainfall-triggered Landslide Susceptibility Mapping and Hazard Warning based on GIS and Machine Learning
Yu et al. Structural integrity estimation of bridges based on acceleration measurement
CN118627407A (en) Waterlogging model creation method and system applied to urban water management

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination